Myocyte-specific enhancer factor 2c triggers transdifferentiation of adipose tissue-derived stromal cells into spontaneously beating cardiomyocyte-like cells

Cardiomyocyte regeneration is limited in adults. The adipose tissue-derived stromal vascular fraction (Ad-SVF) contains pluripotent stem cells that rarely transdifferentiate into spontaneously beating cardiomyocyte-like cells (beating CMs). However, the characteristics of beating CMs and the factors that regulate the differentiation of Ad-SVF toward the cardiac lineage are unknown. We developed a simple culture protocol under which the adult murine inguinal Ad-SVF reproducibly transdifferentiates into beating CMs without induction. The beating CMs showed the striated ventricular phenotype of cardiomyocytes and synchronised oscillation of the intracellular calcium concentration among cells on day 28 of Ad-SVF primary culture. We also identified beating CM-fated progenitors (CFPs) and performed single-cell transcriptome analysis of these CFPs. Among 491 transcription factors that were differentially expressed (≥ 1.75-fold) in CFPs and the beating CMs, myocyte-specific enhancer 2c (Mef2c) was key. Transduction of Ad-SVF cells with Mef2c using a lentiviral vector yielded CFPs and beating CMs with ~ tenfold higher cardiac troponin T expression, which was abolished by silencing of Mef2c. Thus, we identified the master gene required for transdifferentiation of Ad-SVF into beating CMs. These findings will facilitate the development of novel cardiac regeneration therapies based on gene-modified, cardiac lineage-directed Ad-SVF cells.


Global transcriptome analysis of SVF-beating cells.
To investigate the phenotype of the beating SVF cells, we performed a global transcriptome analysis. All of the beating SVF colonies in a dish (~ 100 ng total RNA) on day 28 were manually picked up and combined, constituting the beating-SVF (beating) group. The other cells were collected as the non-beating-SVF (non-beating) group (Fig. 2a,b). Using 10,015 pre-filtered genes, the heatmap image of the microarray expression data showed that the beating group was separated from the non-beating group (Fig. 2c). The class comparison analysis extracted 317 upregulated and 151 downregulated genes in the beating group. Among these genes, pathway analysis revealed several associated with 'direct reprogramming of cardiac fibroblasts into cardiomyocytes' (Fig. 2d). In the beating group, genes associated with sarcomere, ion channels, and transcription factors related to cardiomyocytes as well as skeletal muscle (Fig. 3a) were upregulated, whereas stem-cell-related genes, such as Ly6a and Thy1, were downregulated. Next, we performed a hierarchical clustering analysis of 32,000 genes in adult heart and skeletal muscle tissue and cardiomyocytes isolated from murine embryo or neonatal ventricle (Supplementary Table S2). As shown in Fig. 3b, beating SVF clustered adjacent to neonatal cardiomyocytes and adult cardiac tissue. A scatterplot demonstrated that the gene expression profile of beating SVF was 90.5% similar (< 2.0-fold) to that of neonatal cardiomyocytes,  14. The arrowhead indicates a round cell with intracellular calcium ion oscillation. Arrow shows an elongated cell that is changing from the round form and beginning to beat spontaneously. (g) Maximal rate of shortening (dL/dt) and relaxing (-dL/dt) in each cycle of contractions of a representative isolated SVF beating cell on day 28. (h-j) Synchronised contraction of beating SVF cells on day 28. Phase-contrast image of a cluster of beating SVF cells (e) and the corresponding calcium ion fluorescence image (i). Transient calcium waves of the five cells (j) traced in yellow in panel I (arrows 1 to 5). (k) Transient calcium wave of beating SVF cells with or without ISP. (l) The frequency of calcium ion exchange in beating SVF cells in response to ISP. ISP; isoproterenol. Scale bar, 100 µm. * p < 0.05; *** p < 0.001 by t-test or ANOVA. (m) A representative calcium ion transient trace, represented as ΔF/F 0 , for a single beating SVF cell. (n) A representative action potential of a spontaneously beating SVF that was recorded on day 28 of primary culture. The action potential duration to 50% repolarisation (APD 50 ) was 180 ± 3 and that to 90% repolarisation (APD 90 ) was 277 ± 15 ms. www.nature.com/scientificreports/ and 79.9% to that of adult heart tissue (Fig. 3c). A pairwise Pearson correlation plot showed that beating SVF was highly correlated with neonatal cardiomyocytes (Fig. 3d).
The phenotype of SVF-beating cells was similar to ventricular cardiomyocytes. Because the beating SVF clusters had a similar global gene expression profile to that of neonatal cardiomyocytes, we examined the expression of cardiomyocyte-related genes and the levels of the encoded proteins. qRT-PCR demonstrated that the beating SVF cells expressed genes encoding cardiomyocyte-specific sarcomere proteins, including cardiac muscle troponin T (Tnnt2), myosin heavy chain 6 (Myh6), and actin alpha cardiac muscle 1 (Actc1). The ventricular phenotype of myosin light chain (MLC) 2v (Myl2) was expressed in the beating SVF cells but the atrial type of MLC 2a (Myl7) was not expressed in the beating SVF, as in the non-beating SVF. Likewise, immunocytological analysis demonstrated that the beating SVF cells co-expressed cardiac troponin T (cTnT) with MLC2v (Fig. 4a,b). Co-staining of alpha-actinin with MLC2v indicated the beating SVF cells to be striated muscle with a ventricular phenotype (Fig. 4c). Expression of the pacemaker ion channels encoded by Hcn4 and Hcn2 was increased in beating SVF cells. Expression of the gene encoding secreted brain natriuretic peptide (BNP), but not atrial natriuretic peptide, was enhanced in beating SVF cells, consistent with the expression of ventricular-type MLC. The beating SVF expressed Gja1 at a high level and connexin 43 protein was localised between beating cells (Fig. 4a,d). In the non-beating SVF group, adipocyte-and chondrocyte-specific genes were upregulated (Supplementary Figures S3a and S3c), and some non-beating SVF cells differentiated into adipocytes in the same dish in which beating SVF cells grew (Supplementary Figure S3b).

Analysis of round beating CFPs.
To test our hypothesis that the round cells were the progenitors of CMs, we performed a single-cell-based global gene expression analysis. On day 14 of primary SVF culture, six round cells were picked up using the ALS CellCelector system (Fig. 5a) and combined for microarray analysis. The round cells expressed genes related to cardiac sarcomere components, gap junctions, and pacemaker activity (Fig. 5b). To investigate the key transcription factors for transdifferentiation into CMs, 491 CisBP transcription factors were identified by screening for unchanged genes among the beating, non-beating, and round cells. The heatmap classified these genes into four groups according to their expression pattern: group A, upregulated only in round cells; group B, slightly upregulated in round cells and substantially upregulated in beating cells; group C, downregulated only in round cells; and group D, downregulated in round and beating cells (Fig. 5c,d). Nkx2-5, which is important for differentiation into the cardiac lineage in the embryo, was upregulated only in round cells.

Mef2c was upregulated in round CFPs and in beating CMs.
Because the gene expression profile of beating SVF was suggestive of 'direct reprogramming of cardiac fibroblasts into cardiomyocytes' (Fig. 2d), we examined the expression of the associated transcription factors Gata4, Mef2c, and Tbx5. Gata4, Mef2c, and Tbx5 showed different expression patterns. They were assigned to groups A, B, and D as described above, respectively (Fig. 5c). qRT-PCR confirmed that Gata4 and Mef2c were upregulated in beating SVF, whereas expression of Tbx5 was unchanged, compared to that in non-beating SVF (Fig. 6a). Mef2c was markedly upregulated in beating SVF. Furthermore, immunofluorescence imaging demonstrated that Mef2c and Gata4 were coexpressed in 70% of the round cells (Fig. 6b,d). Moreover, 98% of the beating SVF cells were positive for cardiac troponin T, and 63% of those expressed Mef2c in the nucleus (Fig. 6c,e, and Supplementary Video S10). Since the pathway analysis of the global transcriptome ( Fig. 2d) indicated that the beating SVF clusters expressed skeletal muscleand cardiomyocyte-related genes, we examined the expression of MyoD1, a specific marker of skeletal myocyte differentiation. Flow cytometry demonstrated that on day 14, 9.1% of the SVF cells were positive for MyoD1, and 4.5% were positive for cardiac troponin T (Supplementary Figure S4a). Among the cardiac troponin T-positive cells on day 14, most of which were round cells as confirmed by immunocytochemistry (Supplementary Figure S4c), 81% expressed MyoD1 (Supplementary Figure S4b). The round cells no longer expressed MyoD1 upon becoming elongated beating SVF cells (Supplementary Figure S4c). On day 28, most of the beating SVF cells expressing cTnT were negative for MyoD1 (Supplementary Figures S4a, S4b, and S4d). By contrast, a nonbeating SVF cluster expressing MyoD1 was negative for cTnT (Supplementary Figure S4e).

Overexpression of SVF with Mef2c augmented the yield of round CFPs and their transdifferentiation into CMs.
To examine whether direct reprogramming-related transcription factors regulate the transdifferentiation of SVF into CMs, we transduced SVF with transcription-factor-encoding genes using a lentivirus vector. Overexpression of Mef2c resulted in a 16-fold increase in the number of round cells ( Fig. 7a-d,f) and significantly increased expression of cardiac troponin T (Fig. 7g). Both effects were suppressed by gene silencing using a short hairpin RNA targeting Mef2c (Figs. 7e-g). By contrast, overexpression of Gata4, Tbx5, or control vector had no effect. Although the efficiency of transducing with Mef2c varied among the experiments, expression levels of cardiac troponin T showed a positive linear correlation with those of Mef2c (Fig. 7h). SVF with 40-fold augmented Mef2c gave rise to spontaneously beating CMs with ~ 2000-fold increased cardiac troponin T expression. Therefore, Mef2c positively regulates the transdifferentiation of SVF into CMs.

Discussion
We identified simple culture conditions under which murine inguinal Ad-SVF reproducibly and spontaneously transdifferentiated into beating CMs without induction. This phenomenon occurred only in primary SVF culture and was mediated by round cells. We evaluated the global gene expression profiles of the spontaneously differentiated beating CMs and the round progenitor cells. We also identified the master gene required for SVF transdifferentiation into beating CMs.  20 . The mechanism of action of MSCs in cardiac regeneration therapy is protection of injured cardiomyocytes by secreted cytokines and growth factors. Several failed attempts have resulted in a consensus that MSCs do not differentiate into beating CMs 9,21,22 . Bone-marrow-derived stromal cell lines can be induced to differentiate into spontaneously beating CMs by 5-azacytidine 16 . Subcutaneous Ad-SVF shows promise for clinical applications because the need for expansion in culture 13,15 is removed by the abundant multipotent stem cells that can transdifferentiate into adipocytes, chondrocytes, or cardiomyocytes 23,24 . Planat-Bénard et al. reported transdifferentiation of Ad-SVF into beating CMs using semi-liquid medium and supplementation with cytokines and growth factors 17 . The use of semisolid culture with enzyme-crosslinked gelatine hydrogels obviates the need for supplements to induce differentiation of SVF into beating CMs 18 . By contrast, we achieved differentiation of SVF into spontaneously beating CMs using an uncoated culture dish and standard liquid medium supplemented only with a reducing agent. By minimising external modification of SVF cells, their original characteristics when undergoing transdifferentiation into CMs were revealed, without reducing the efficiency of beating SVF cell production. In this study, we performed immunofluorescence imaging, intracellular calcium imaging, electrophysiological recording, and single-cell-based global gene expression profiling of beating SVF clusters.
Expression profiling of beating SVF clusters was hampered by the presence of non-beating SVF. In this study, beating SVF clusters were manually selected and RNA was extracted for global gene expression profiling. Consistent with previous reports 17, 18 , the beating SVF clusters expressed genes encoding cardiac sarcomere, gap junction, pacemaker ion channel, and secretory proteins. The levels of the corresponding proteins (cardiac troponin T, ventricular myosin light chain, and connexin 43) were confirmed by immunofluorescence. The global gene expression profiles of the beating SVF clusters were comparable to the profiles of neonatal cardiomyocytes 25,26 but differed from those of adult cardiomyocytes. In an electrophysiological study, the beating CMs had a prolonged action potential with a small plateau phase. Consistently, the intracellular calcium influx decreased gradually in the systolic phase, similar to previous studies of a rodent ventricular phenotype 27 . Collectively, the beating SVF cells were similar to cardiomyocytes.
The factor that regulates the switch to the cardiomyocyte lineage during SVF primary culture is important. In adults, cardiomyocytes are terminally differentiated and do not grow 14 . Transcription of cardiomyocyterelated genes is strictly regulated by epigenetic processes including CpG methylation, histone modifications, and noncoding RNAs [28][29][30][31] , which restrict the activation of transcription factors related to cardiac differentiation 29 . A pathway analysis of the microarray data showed that the expression of genes related to 'direct reprogramming of cardiac fibroblasts into cardiomyocytes' was altered in beating SVF cells. Direct reprogramming shows promise for regeneration therapy 26 , in which application of cardiac fibroblasts in combination with transcription factors-including Gata4, Mef2c, and Tbx5-induces transformation into cardiomyocytes 32 . Indeed, in the beating SVF group, Gata4 was upregulated and Mef2c was highly expressed; by contrast, expression of Tbx5 was unchanged. Additionally, in the non-beating SVF group, adipocyte-and chondrocyte-specific genes were upregulated. Therefore, reprogramming of the cells occurred during the primary culture of SVF, and the reprogrammed cells differentiated into cardiomyocytes, as well as adipocytes or chondrocytes. The ultimate goal of the Ad-SVF-derived CMs is to selectively generate abundant functional CMs for cardiac regeneration applications. The addition of supplements including cytokines and growth factors, as used previously 17 , did not increase the yield of beating CMs. However, we found that the transcription factor Mef2c was crucial for activating SVF cells towards cardiac differentiation. Transduction of the SVF cells with Mef2c using a lentiviral vector allowed sustained augmentation of gene expression during cell multiplication, which increased CM generation. Notably, the number of beating CMs was directly proportional to Mef2c expression. Therefore, our results indicate that primed SVF cells transduced with Mef2c have therapeutic potential in heart failure.
The fact that the beating SVF cells originated from round cells led to the hypothesis that the latter were CFPs. The round cells were readily discriminated from other cells by phase-contrast microscopy, enabling isolation of-single cells 33 and assessment of their global gene expression profiles. The expression of transcription factors could be classified into four groups, and some cardiomyocyte-specific genes were upregulated even in round cells. Also, Mef2c was upregulated in round cells and this was retained in beating SVF; by contrast, upregulation of Gata4 and Nkx2-5 was transient. These results imply that the round cells were CFPs. The CFPs expressed specific genes, including those of cell-differentiating antigens (data not shown), which are candidate markers of CFPs. Note, however, that the CFP transcriptome in this study was the result of one round of the experiment based on a pooled sample of six hand-picked cells, raising the concern that noise was introduced; we hope to verify this in our next study by using more samples. in the beating compared to the non-beating group among 468 genes extracted by class comparison analysis (p < 0.0001) from 10,015 prefiltered genes (FDR: 10 −7 to 0.002). FDR, false discovery rate. (b) Hierarchical clustering analysis of 32,000 genes (genes with a log-ratio variation in the 10th percentile were excluded) using an unsupervised learning method, compared with the SVF group based on array (Affymetrix Mouse Gene 1.0 ST Array) data from the gene expression omnibus database (NIH ; Table S2). (c) Scatterplot of gene expression [log 2 (intensity)] between beating SVF and adult heart or neonatal cardiomyocytes. Genes with a ≤ twofold difference in expression are shown in red and the ratios to all genes are indicated. (d) Image plot of pairwise Pearson correlation of expression of arrays of beating SVF comparing to neonatal cardiomyocytes, embryo cardiomyocytes, adult heart, and skeletal muscle. The arrays are ordered based on hierarchical clustering using the average linkage method. Colours indicate the correlation coefficients. The graphical images of hierarchical clustering (b), scatter plot (c), and image plot (d) were generated using BRB Array Tools software (v. 4.6.0) (NIH, Bethesda, MD, USA) (URL: https ://brb.nci.nih.gov/BRB-Array Tools /). www.nature.com/scientificreports/ www.nature.com/scientificreports/ Our results also raise several questions. First, the global transcriptome of the beating SVF group represented a mixed cardiomyocyte/skeletal myocyte phenotype. On day 28, when the microarray analysis was performed, immunofluorescence staining revealed that the beating SVF cells all expressed cardiac troponin T, but not MyoD1, which is a specific transcription factor regulating skeletal muscle differentiation 34 . By contrast, a nonbeating cell cluster expressing MyoD1 was negative for cardiac troponin T. Therefore, we considered a beating SVF cell to be a cardiomyocyte, but not a skeletal myocyte. As we selected the beating SVF clusters manually, we speculated that they contained not only beating SVF cells but also round cells (mostly expressing MyoD1) and nearby non-beating skeletal myocytes. We postulated that was why the transcriptome of the beating SVF group represented a mixed cardiomyocyte/skeletal myocyte phenotype. Using an immunofluorescence approach, the round cells considered to be CFPs exhibited transient MyoD1 expression; subsequently, the CFPs no longer expressed MyoD1 upon becoming cTnT-positive beating SVF cells. Although MyoD1 is a master gene involved in skeletal muscle 34 , its expression was induced under hyperoxic conditions in cultured chick foetal cardiac myocytes 35 . Furthermore, fusion of the MyoD domain to Mef2c, Gata4, Hand2, and Tbx5, increased the efficiency of direct reprogramming of fibroblasts into cardiomyocytes 36 . From these observations, we speculate that the transient expression of MyoD1 in round cells plays a crucial role in their differentiation into cardiomyocytes. The mechanism underlying the regulation of differentiation of SVF cells into beating cardiomyocytes or non-beating skeletal myocytes remains unclear. However, the duration of MyoD1 expression might be a key. We will address this using a time-lapse tracing of CFPs or MyoD1-fate mapping in our next study. We also evaluated miRNA expression in the beating SVF and round CFPs (data not shown). miRNAs have the potential to modulate the differentiation of stem cells [37][38][39] . Thus, identification of the key microRNA that regulates the differentiation of CFPs against skeletal myocytes but toward cardiomyocytes is considered a reasonable approach in the development of "pure" cardiomyocytes 40 , which will be investigated in our next study. Second, it is of note that spontaneous cardiac differentiation occurred in SVF primary culture but not in subcultured cells. Li and co-workers reported that Gata4, Tbx5, and Baf60c induced differentiation of repeated cultured adipose-derived stromal cells into CMs, albeit on a small scale 41 . Because the phenotype of MSCs changes during subculture 42 , additional key factors other than Mef2c that are essential for spontaneous differentiation into beating CMs but are lost in the course of repeated subcultures should also be identified. Third, the significance of Tbx5 in the differentiation of SVF cells into beating CMs in this study was unclear. Tbx5 is a key transcription factor regulating cardiac differentiation and is essential for direct remodelling of fibroblasts into cardiomyocytes 26 . The role of Tbx5 in SVF cell differentiation could be evaluated by its knockdown, which we will perform in our next study. Fourth, the beating CMs had a ventricular, rather than an atrial, phenotype and expressed MLC2v and BNP. According to a previous report 43 , during embryonic development of the four-chambered heart, expression of MLC2v is regulated by Mef2c and Nkx2-5 and represses ANF expression in the ventricle. Further investigation of the regulation of differentiation of SVF into chamber-specific cardiomyocytes is needed. Finally, this was an in vitro study. In previous clinical trials, intramyocardial injection of skeletal myoblasts failed to improve heart function due to ventricular arrhythmia caused by electrical uncoupling 44 . Although the beating CMs expressed connexins 40 and 43 and demonstrated synchronised contraction, whether the beating CMs are functional in a transplanted heart in vivo needs further study. Furthermore, the administration of Mef2c-primed SVF or augmented CMfated round cells in an animal model of heart failure is needed to investigate their safety (arrhythmogenicity) and efficacy (engraftment and improving ventricular function). In conclusion, we identified the master genes required for Ad-SVF transdifferentiation into beating CMs. The findings will enable the development of novel cardiac regeneration therapies using gene-modified, cardiac lineage-directed Ad-SVF.

Methods
All experiments were carried out in accordance with approved guidelines and were approved by the Animal Experiment Committee at the Institute for Experimental Animals, Kanazawa University.

Isolation of Ad-SVF. Bilateral inguinal subcutaneous adipose tissues without lymph nodes, weigh-
ing ~ 500 mg, were obtained from a C57BL/6J mouse (aged 8-12 weeks; male or female). After washing with sterile phosphate-buffered saline (PBS), the tissues were digested using the Tumor Dissociation Kit, Mouse (Miltenyi Biotec, Bergisch Gladbach, Germany) in serum-free Dulbecco's modified Eagle's medium (DMEM)/F-12 (Thermo Fisher Scientific, Waltham, MA, USA) for 30 min at 37 °C using gentleMACS (Miltenyi Biotec). The digested tissue was thoroughly pipetted ~ 10 times in DMEM supplemented with 10% foetal bovine serum (FBS) (F7524, Sigma-Aldrich, St. Louis, MO, USA) and filtered through a 100-µm nylon mesh, followed by centrifugation at 400× g for 5 min. After washing with DMEM/F-12 with FBS and centrifugation, the cell pellet was resuspended in BD Pharm Lyse (BD Biosciences, Franklin Lakes, NJ, USA) and incubated for 5 min to lyse red blood cells. After a further round of centrifugation, the cell pellet was filtered through a 35-µm mesh and centrifuged at 400× g for 5 min; then, the SVF was resuspended in culture medium. ASCs. To obtain ASCs, the SVF cells were passaged at sub-confluence, and those from passage 3 were used as ASCs.  Action potential recordings. The action potential of the beating SVF cells was recorded on day 28, as described previously 45 . Briefly, the action potential was recorded using patch pipettes at 37℃ with an external control solution containing (in mmol/L) NaCl 140, KCl 5.4, CaCl 2 1.8, MaCl 2 2, glucose 10, and HEPES 10 (pH 7.4). Suction pipettes were generated from borosilicate capillary tubes (8250 glass; A-M Systems) and firepolished to obtain a resistance of 5-8 MΩ when filled with a solution containing (in mmol/L) KCl 120, EGTA 5, K 2 ATP 5, MgCl 2 2.5, and HEPES (pH 7.2). Under the Axiovert 10 (Carl Zeiss, Jena, Germany) microscope, Vm was measured using the AxoClamp 2A amplifier (Molecular Devices) in bridge mode via the disrupted patch technique, as described previously 43 . Vm was digitised using the PowerLab system (AD Instruments, Dunedin, New Zealand) and recorded at a sampling frequency of 2 kHz/16-bit in LabChart 7.0 (AD Instruments) using a 0.5-40 Hz band-pass filter. All recording equipment was housed on an air table within a grounded Faraday cage to minimise background noise. The APD 50 and APD 90 of repolarisation were calculated from a series of 10 potentials.

Assessment of contraction and relaxation.
Raw phase-contrast video files of spontaneously beating cells were imported into Kinovea v. 0.8.15. A small region of interest (ROI) on the beating CMs was defined as the tracking point. Acceleration in each frame at the point of interest was extracted using the software in semiautomatic mode, wherein a tracking point is placed and the software attempts to automatically track those pixels into the next frame. The maximum rates of contraction and relaxation for each cycle were determined by extracting the acceleration data in a frame-by-frame manner.
Pharmacological study. A fluorescence video of beating CMs of interest was obtained through intracellular calcium imaging, and a calcium transient wave was observed. Next, the frequency of calcium ion exchange was calculated. The chronotropic response was assessed by adding 0.2 to 10 µM isoproterenol (Sigma-Aldrich) to CMs.

Extraction of RNA from beating cell clusters.
Beating cell clusters were manually picked from a 100mm dish using tweezers (MA50, Dumond, Switzerland) under an inverted optical microscope (CKX41, Olympus, Tokyo, Japan) at 40-100 × magnification, and then transferred to rinsing buffer (Buffer RA1 added to Tris (2-carboxyethyl) phosphine [TCEP]; NucleoSpin RNA XS [Merchery-Nagel, Düren, Germany]). The beating cell colonies obtained from two or three dishes derived from one lot of SVF were combined into one sample, which was processed for RNA extraction without addition of carrier RNA, according to the manufacturer's  Picking single round cells. Single round cells were picked using an ALS CellCelector (ALS, Jena, Germany). The CellCelector combines an automated high-content imaging cytometry platform with automated cell micromanipulation using a vertical high-precision glass micro-capillary fixed to a precision robotic arm and a syringe pump. Briefly, scanning of round cells of interest in a phase-contrast channel was followed by automated identification of the target cells and recovery into a destination vessel (PCR tube) for downstream analysis.
Microarray analysis. Preparation of RNA and microarray hybridisation. For microarray analysis, the quality of RNA was confirmed (RNA integrity number > 8) using a bioanalyser (RNA 6000 Nano Assay, Agilent Technologies). Total RNA (50 ng) was processed to generate sense-strand cDNA (ss-cDNA) using the Ovation Pico RNA WTA System (Tecan Genomics, Maennedorf, Switzerland). Biotin-labelled RNA or DNA fragments were hybridised to GeneChip Mouse Gene 1.0 ST Arrays (Affymetrix, Santa Clara, CA, USA) and quantified. Images were processed and cell intensity (CEL) files were generated using GeneChip Command Console software (Affymetrix).
Data analysis. Raw data were imported into BRB Array Tools software (v. 4.6.0) (NIH, Bethesda, MD, USA) (URL: https ://brb.nci.nih.gov/BRB-Array Tools /) and log 2 -scaled, normalised-mean-centred, and subjected to average-linkage clustering. Hierarchical clustering of both gene expression (rows) and the independent samples (columns) in the data matrix was performed. The rows/columns of the data matrix were re-ordered such that the most similar expression patterns or samples were located adjacent to each other. To identify genes that were differentially expressed among the groups, the class comparison tool, which is based on univariate t-tests, of BRB-Array Tools was used. Gene Ontology analysis was performed using the functional ontology enrichment tool of MetaCore v. 19.4 (Clarivate Analytics, Tokyo, Japan). Data for genes that were differentially expressed among the groups were imported with the corresponding fold-changes and p values.
Quantitative PCR. Total RNA (100 ng) was used to generate cDNA using TaqMan Universal Master Mix.
Quantitative real-time polymerase chain reaction (qRT-PCR) was performed using an ABI Prism 7300 Sequence Detection System (Applied Biosystems, Foster City, CA, USA). Beta-actin (Mouse ACTB, 4352341E) was used as an endogenous control. Relative expression was assessed using the comparative CT method, correcting for the amplification efficiency of the primers, and each sample was analysed in duplicate (TaqMan assay IDs are listed in Supplementary Table S1).